Passing Efficiency Beyond Passer Rating: YPA, ANY/A, CPOE, and EPA

The NCAA passer rating is a fine headline number and a terrible final verdict. I've already walked through exactly how that formula works — four box-score inputs, four weights, one division — so I won't re-derive it here. The point of this piece is the next step: once you accept that passer rating is blind to down, distance, and opponent, what do you reach for instead? There's a whole ladder of passing-efficiency measures, each one adding back a piece of context the rating throws away. Let me climb it rung by rung — yards per attempt, adjusted net yards per attempt, completion over expected, and EPA per dropback — and say plainly what each one buys you and where it still falls short.

The one-sentence version: every metric above passer rating is an attempt to answer "yards for what, against whom, in what situation" — and the further up you go, the more of the real game the number can see.

Yards per attempt: the honest baseline

If you only get one passing number, make it yards per attempt — total passing yards divided by pass attempts. It sounds primitive, but it's the single most predictive box-score passing stat, and it quietly sits inside passer rating already (the 8.4-per-yard term is just yards-per-attempt in disguise). Why does it work so well? Because it rewards the thing that actually moves drives — gaining ground per throw — without the distortions of raw volume. A quarterback can pile up 400 yards by throwing 55 times; yards per attempt asks whether each of those throws was worth taking.

Its blind spots are obvious, though. It treats a 9-yard completion on 3rd-and-12 (a failure) the same as a 9-yard completion on 3rd-and-3 (a conversion). It says nothing about interceptions — a quarterback can average a gaudy figure while throwing the ball to the other team. And it ignores sacks entirely, which is a serious omission, because a sack is a failed dropback that the passer often had a hand in. So yards per attempt is the honest baseline: better than total yards, better than completion percentage alone, but still missing turnovers, sacks, and situation.

ANY/A: folding in touchdowns, picks, and sacks

Adjusted net yards per attempt — ANY/A — is the natural fix for two of those holes. It starts from yards per attempt and bolts on a bonus for touchdowns, a penalty for interceptions, and, crucially, the sack yardage that plain passing stats ignore. The standard construction looks like this:

ANY/A = (Pass Yards + 20×TD − 45×INT − Sack Yards) / (Attempts + Sacks)

Read it slowly and the philosophy is clear. A touchdown is worth a flat 20-yard bonus, an interception costs 45 yards — a heavy penalty, because turnovers are drive-enders — and sacks appear in both the numerator (you lose the yards) and the denominator (a sack counts as a failed dropback alongside the attempts). That denominator change is the whole point: ANY/A grades a quarterback per dropback, not per pass, so a passer who takes a lot of sacks can't hide them by simply not throwing. The specific weights here are the widely used public-analytics values; treat them as a sensible convention, not a law of nature.

What does ANY/A buy you over passer rating? It speaks in yards, which is intuitive, and it explicitly charges for sacks, which passer rating cannot see at all. Where it still falls short: like everything so far, it's blind to down and distance, and it doesn't adjust for opponent. A fat ANY/A built against soft defenses is still flattering.

Completion over expected: crediting the difficulty of the throw

Raw completion percentage is nearly useless on its own, because not all throws are equally catchable. A checkdown to a running back in the flat and a back-shoulder dart into tight coverage 25 yards downfield both count as one completion, but they are wildly different feats. Completion percentage over expected (CPOE) tries to fix that by first estimating how likely an average passer would be to complete each throw — based on factors like air yards, the distance to the nearest defender, and where the receiver is on the field — and then measuring how far above or below that baseline a quarterback actually performs.

So if a model says a given set of throws should be completed 64% of the time and the quarterback hits 70%, his CPOE is +6 percentage points: he's completing passes a typical passer would miss. This is the first metric on the ladder that tries to grade accuracy independent of scheme — a quarterback in a dink-and-dunk offense won't get artificially rewarded just because short throws are easy to complete, and a downfield gunslinger won't get punished for attempting hard throws.

The honesty caveat matters more here than anywhere else: CPOE depends entirely on the expected-completion model behind it, and those models are built by different analysts from tracking data, so the exact numbers vary by source and aren't something you can compute from a box score. Treat CPOE as a directional, model-dependent accuracy signal, not a precise universal constant. Its other limit is that it's purely about completing passes — it doesn't know whether the completion gained enough to matter, or whether an incompletion was actually the smart throw-away.

EPA per dropback: the situational summit

At the top of the ladder sits expected points added per dropback. I've explained EPA from scratch elsewhere, so briefly: every game situation — down, distance, field position — has an expected-points value learned from thousands of historical drives, and EPA is simply how much a play changed that value. Apply it to a quarterback's dropbacks (passes and sacks and scrambles) and average it, and you get EPA per dropback: the closest thing we have to a single, situation-aware grade of passing value.

This is the only metric on the ladder that fully respects the situation. A 6-yard completion on 3rd-and-3 produces positive EPA because it converted; the same 6 yards on 3rd-and-12 produces negative EPA because the drive likely stalled. A deep ball that flips field position is rewarded for the field position, not just the yardage. Sacks and interceptions show up as the large negative swings they truly are. Where passer rating asks "what did the box score say," EPA per dropback asks "how much did each dropback actually help the team score," which is the question that was worth asking all along.

Its limits are real but different in kind. EPA per dropback needs play-by-play data and an expected-points model, so you can't scribble it on a napkin. It still doesn't adjust for opponent unless you layer that on separately — raw EPA padded against a bad defense is the advanced-stats version of empty yards. And like any per-play efficiency number, it's noisy over small samples; one busted coverage giving up a 70-yard touchdown can swing a single game's figure hard. It's a season-scale stat.

Where passer rating actively misleads

Pulling it together, here are the specific situations where leaning on passer rating will steer you wrong, and which rung of the ladder to grab instead.

• The checkdown merchant. A quarterback in a quick-game offense can post a glittering passer rating on high completion percentage and short throws, because the formula's completion and touchdown bonuses are generous. Yards per attempt and EPA per dropback will both deflate that profile toward reality.
• The sack-prone passer. Passer rating can't see sacks, so a quarterback who avoids interceptions by holding the ball and eating sacks looks cleaner than he is. ANY/A and EPA per dropback both charge him for the lost-down dropbacks.
• The garbage-time accumulator. Yards and touchdowns piled up in blowouts inflate passer rating just as much as meaningful production. Situation-aware EPA helps, and an opponent or leverage adjustment helps more.
• The volume illusion. Because passer rating is a per-attempt rate, a passer can look efficient while contributing little if his efficient throws are all low-leverage. Down-aware measures catch the difference between efficient and useful.

None of this means passer rating is garbage. It's a clean, transparent, universally available summary — a perfectly good first glance. The mistake is treating a first glance as the last word. Each rung above it — yards per attempt for an honest baseline, ANY/A to fold in turnovers and sacks, CPOE for scheme-independent accuracy, EPA per dropback for situation — buys back a slice of context the rating discarded, at the cost of a little more data and a little less napkin-friendliness.

How to use the ladder

• Start with yards per attempt. If you only have a box score, it's the most honest single number, and it beats both raw yards and completion percentage.
• Reach for ANY/A when sacks and turnovers matter. It's still box-score computable and it stops a passer from hiding sacks by not throwing.
• Use CPOE as an accuracy lens, not a verdict. It's model-dependent and source-dependent — great for "is this guy accurate beyond his scheme," not for precise cross-source comparison.
• Trust EPA per dropback over a season. It's the most complete picture, but it's noisy in small samples and unadjusted for opponent unless you say otherwise.
• Never let one number be the verdict. Read two or three rungs together, and remember which context each one still can't see.

The passer rating tells you what happened in the box score. The rest of the ladder tells you what it was worth. Climb as far as your data allows, label your hypotheticals honestly, and you'll never again mistake a pretty rating for a good quarterback.